NumPEx Transversal Workgroups

Focuses on harnessing accelerators (such as GPUs and specialized hardware) and exploring programming models suited for heterogeneous, high-performance architectures. Key objectives include: - Evaluating performance and portability across CPU-GPU-FPGA platforms - Developing best practices for programming models (MPI+X, OpenMP, etc.) - Addressing code modernization and tooling for accelerated computing

Addresses the intersection of Artificial Intelligence (AI) and High-Performance Computing (HPC), encompassing: - AI for HPC: leveraging ML/AI techniques to enhance HPC workflows (e.g., surrogate modeling, code auto-tuning) - HPC for AI: accelerating large-scale AI tasks and deep learning workloads - Cross-pollination of methods and shared infrastructure among NumPEx projects

Focuses on the coordination and optimization of national and regional computing centers, aiming to: - Streamline resource allocation for large-scale NumPEx simulations - Foster infrastructure interoperability and distributed job scheduling - Develop joint strategies for data management and transfer

Investigates methodologies to measure, optimize, and reduce energy consumption in HPC: - Metrics and models for energy usage - Energy-aware runtime scheduling and resource allocation - Hardware-software co-design to improve energy efficiency

Aims to ensure reliability and fault tolerance on emerging exascale systems: - Checkpoint/restart strategies and advanced fault-tolerant techniques - Silent error detection and correction - Redundancy and rollback mechanisms

Coordinates NumPEx’s educational initiatives, including: - Courses, tutorials, and workshops on HPC, exascale methods, and specialized topics - Dissemination of best practices in software development - Fostering the next generation of HPC professionals

Focuses on building robust software engineering practices across NumPEx: - Continuous integration/continuous deployment (CI/CD) pipelines - Shared libraries, packaging systems, and deployment strategies - Ensuring interoperability among various HPC components

Ensures diversity and inclusion principles are integrated throughout NumPEx: - Developing policies and guidelines for equitable collaboration - Organizing events and discussions that promote diversity in the HPC community - Tracking and improving representation metrics within projects

Strengthens ties with global HPC initiatives and research partners: - Identifying key international joint projects or calls - Facilitating researcher exchanges and shared resources - Aligning NumPEx with worldwide exascale strategies